eval optimizations for emission #5
Description
Using eval violates strict CSP policy. However, it can bring huge perf benefits at the same time. I wonder, is it possible to use CSP and allow usage of eval, say, for current domain only.
My tests show that using eval technique for compose function provides almost the same perf as by writing function composition by hand. (web) eval-compose is slower by 5%, while reduce-compose is slower x3 times.
If it would be possible to apply to the graph computation as an optimization on per-graph basis (similar to current graph cache, but inline a lot more) it can yield big perf gain. Keep in mind that it would cause a bigger memory footprint and still worth to measure how much memory would be taken in that case.